As noticed by Android Authority on Monday morning, some YouTube users are now seeing a disruptive prompt when they try to watch videos while using a browser with an ad blocker. The prompt notes that "ad blockers violate YouTube's Terms of Service" and urges users to either allow ads on YouTube or subscribe to YouTube Premium.

YouTube tries to kill ad blockers in push for ad dollars, Premium subs

Airbyte announced its "Powered by Airbyte" version that enables makers of software to embed over 100 integrations into their applications.

Powered by Airbyte Offers Software Makers More than 100 Data Integrations

This car owner redefined “search engine” when he used Google Earth to track down and recover both of his vehicles after they were pilfered from his driveway.

I used Google Earth to track down the thieves who stole my cars

Nick Clegg, Meta's president of global affairs, spoke ahead of the landmark AI Summit being hosted at Bletchley Park in the U.K. The summit is expected focus on mitigating the potential harms of AI.

Meta exec and former U.K. Deputy Prime Minister compares AI fears to past ‘moral panic’ over video games—and bicycles

The U.S. antitrust lawsuit against Alphabet ( NASDAQ: GOOG) ( NASDAQ: GOOGL) is starting to get to Barclays. “We’ve tracked the DoJ trial closely over the past several weeks and are growing increasingly concerned about the strength of the case against Google,” analysts led by Ross Sandler wrote in a note.

Antitrust suit against Google making Barclays 'increasingly concerned'

Available on Google Cloud Marketplace and Will Integrate with Vertex AI to Help Customers Process and Analyze Data Box, Inc. (NYSE:BOX), the leading Content Cloud, and Google Cloud today announced an

Box and Google Cloud Expand Strategic Partnership Across Generative AI and Go-to-Market

Alphabet experienced strong revenue growth during the pandemic but has started to normalize in the post-pandemic era. Read more about GOOG stock analysis here.

Google: Advertising Revenue Starts To Normalize, Initiates With 'Buy'

DeepMind co-founder Demis Hassabis hit back against criticisms from Meta Platforms Inc.’s chief AI scientist — who had criticized him and others in the industry for playing up the existential risks of the technology — calling them preposterous.

Google Deepmind Chief Calls Meta’s AI Criticisms ‘Preposterous’

In our experience at American Consumer Claims, we have never heard of a client proactively approaching a timeshare company intending to purchase. Perhaps it has happened, but the vast majority of

Escaping the $10.5 billion U.S. timeshare trap

The Microsoft 365 Copilot tool will appear in Word, Excel and other Office programs and cost $30 per person per month.

Microsoft starts selling AI tool for Office, which could generate $10 billion a year by 2026

DeepMind

deepmind

Today, I received none other than on my podcast, a trailblazer in AI product management and a visionary who’s been at the helm of transformative projects at and , working on the most advanced LLM projects. Paige Bailey Google DeepMind GitHub In our discussion, Paige shares her insights on the current state and the potential of . We touch upon Google’s PaLM 2 and its implications for the future of machine learning. generative AI Here are some topics and insights we touched on in the discussion: Paige emphasizes the transformative power of technology to make experiences once exclusive to the ultra-wealthy accessible to all, such as the services provided by Uber and Instacart. Discusses the challenges and advancements in AI-assisted coding and the role of GitHub Copilot in shaping developer experiences. Highlights the importance of careful product design around large language models to reduce errors and enhance performance, citing Google's approach with models like Bard. Underscores the need for responsible AI development, focusing on safety checks and ethical considerations in technology deployment. Paige points out the revolutionary impact of foundational models in inspiring businesses to embed research teams for innovation. Notes the potential risks and benefits of using AI in educational settings, particularly regarding the reliance on AI for problem-solving in academic learning. She also provides insight into the role of a product manager within AI research teams, focusing on prioritizing use cases and data for model training. If these topics sound interesting, join us in this discussion and listen on , or now: Spotify Apple podcast YouTube https://youtu.be/y5nxMnTtbD4?embedable=true&transcript=true

Paige Bailey: Pioneering Generative AI in Product Management at Google DeepMind

In this captivating podcast episode, I had the opportunity to interview Aleksa Gordić, an ex-research engineer at who ventured into creating his own startup called Ortus. The interview spans a range of topics, shedding light on Aleksa's current priorities, his work at DeepMind, and his decision to leave academia and pursue his own entrepreneurial path. DeepMind The podcast begins with a discussion about Aleksa's current focus and his commitment to both Ortus and his YouTube channel, AI Epiphany. Despite managing multiple projects, Aleksa expresses his dedication to continue creating content and potentially delve back into academic research. The conversation then shifts to his experience at DeepMind, where he worked as a research engineer in machine learning, primarily in the field of computer vision. Addressing Aleksa's educational background, we explore his decision to drop out of his master's program and the reasons behind it. He explains how he recognized the need for a more practical and hands-on approach, which led him to pursue software engineering roles. Surprisingly, Aleksa successfully landed a software engineer position at Microsoft without an official degree in machine learning, showcasing the significance of practical skills and experience in the field. Throughout the interview, Aleksa provides valuable insights into the process of securing AI-related jobs at renowned companies like DeepMind and Microsoft. He emphasizes the importance of continuous learning, engaging in competitions such as Kaggle, and building a strong portfolio through projects and open-source contributions. The conversation then transitions to Aleksa's decision to leave DeepMind and embark on his entrepreneurial journey with Ortus. He discusses the challenges he faced, the unique value proposition of Ortus—a Google Chrome extension that uses large language models to answer specific video-related questions—and how his technical expertise aided in the creation of the platform (basically allowing you to discuss with YouTube videos, including mine!). The interview touches on various other topics, including the future of AI in content creation, productivity-enhancing resources and tips, and the benefits of physical training. Throughout the discussion, Aleksa shares valuable insights and personal experiences, providing inspiration and guidance for individuals interested in pursuing a career in AI. I highly recommend listening to the complete podcast episode on , , or to learn from Aleksa's experiences, challenges, and vision for the future of AI. Don't miss out on this opportunity to explore the dynamic world of AI research and entrepreneurship firsthand. Tune in now and be part of the AI revolution! YouTube Spotify Apple Podcasts https://youtu.be/KdyAE1vhLmY?embedable=true&transcript=true

Insights from a DeepMind Researcher Turned Startup Founder

Perspectivas de un investigador de DeepMind convertido en fundador de una empresa emergente

Perspectives d'un chercheur de DeepMind devenu fondateur d'une startup

एक डीपमाइंड शोधकर्ता से स्टार्टअप संस्थापक की अंतर्दृष्टि

DeepMind 研究者からスタートアップ創設者に転身した人物からの洞察

Insights de um pesquisador da DeepMind que se tornou fundador de uma startup

Thông tin chi tiết từ Nhà nghiên cứu DeepMind trở thành Nhà sáng lập Startup

DeepMind 研究员转型为初创公司创始人的见解

Gato from DeepMind was just published! It is a single transformer that can play Atari games, caption images, chat with people, control a real robotic arm, and more! Indeed, it is trained once and uses the same weights to achieve all those tasks. And as per Deepmind, this is not only a transformer but also an agent. This is what happens when you mix Transformers with progress on multi-task reinforcement learning agents. As we said, Gato is a multi-modal agent. Meaning that it can create captions for images or answer questions as a chatbot. You’d say that GPT-3 can already do that, but Gato can do more… The multi-modality comes from the fact that Gato can also play Atari games at the human level or even do real-world tasks like controlling robotic arms to move objects precisely. It understands words, images, and even physics... Learn more in the video References ►Read the full article: ►Deepmind's blog  post: ►Paper: Reed S. et al., 2022, Deemind: Gato, ►My Newsletter (A new AI application explained weekly to your emails!): https://www.louisbouchard.ai/deepmind-gato/ https://www.deepmind.com/publications/a-generalist-agent https://storage.googleapis.com/deepmind-media/A%20Generalist%20Agent/Generalist%20Agent.pdf https://www.louisbouchard.ai/newsletter/ Video transcript 0:00 Gato from deepmind was just published 0:02 it's a single transformer that can play 0:04 atari games caption images chat with 0:07 people control a real robotic arm and 0:09 more indeed is trained once and uses the 0:12 same weights to achieve all those tasks 0:15 and as per deepmind this is not only a 0:17 transformer but also an agent this is 0:20 what happens when you mix transformers 0:22 with progress on multi-task 0:23 reinforcement learning agents as we said 0:26 gato is a multi-modal agent meaning that 0:29 it can create captions for images or 0:31 answer questions as a chatbot you'd see 0:34 that gpt3 can already do that but ghetto 0:36 can do more the multimodality comes from 0:39 the fact that ghetto can also play atari 0:41 games at the human level or even do real 0:44 world tasks like controlling robotic 0:46 arms to move objects precisely it 0:48 understands words images and even 0:51 physics ghetto is the first generalist 0:54 model that performs so well on so many 0:56 different tasks and it's extremely 0:58 promising for the field it was trained 1:00 on 604 distinct tasks with varying 1:03 modalities observations and action 1:06 specifications making it the perfect 1:08 generalist and as i said it does all 1:11 that with the same network and weights 1:13 and before you ask it only needs 1.2 1:15 billion parameters compared to gpt3 that 1:18 requires 1:19 175 billion of them it's not a trap 1:22 where you have to retrain or fight unit 1:24 to all tasks you can send both an image 1:27 and text and it will work you can even 1:29 add in a few movements from a robot arm 1:32 the model can decide which type of 1:34 output to provide based on its context 1:36 ranging from text to discrete actions in 1:38 an environment if you enjoyed the video 1:41 please consider subscribing and let me 1:43 know if you like this kind of news video 1:46 i definitely do more this is possible 1:48 because of their tokenization process 1:50 tokenization is when you prepare your 1:52 inputs for the modal as they do not 1:55 understand text or images by themselves 1:57 language models and ghetto took the 1:59 total number of sub words for example 32 2:02 000 and each word has a number assigned 2:05 to it for images they follow the vit 2:08 patch embedding using a widely used 2:10 resnet block as we covered in a previous 2:12 video we also tokenized the button 2:14 presses as integer numbers for atari 2:16 games or discrete values finally for 2:19 continuous values like proprioceptive 2:21 inputs we talked about with the robotic 2:23 arms they encoded the different track 2:25 matrix into float numbers and added them 2:27 after the text tokens using all those 2:30 different inputs the agent adapts to the 2:32 current task to generate appropriate 2:34 outputs during training they use prompt 2:36 conditioning as in gpt3 with previously 2:39 sampled actions and observations the 2:42 progress in generalist rl agents in the 2:44 last years has been incredible and came 2:47 mainly from deepmind one could see that 2:49 they are moving the needle closer to 2:51 general ai or human level intelligence 2:55 if we can finally define it i love how 2:57 many details they gave in their paper 2:59 and i'm excited to see what they will do 3:01 or what other people will do using this 3:03 model's architecture the link to the 3:06 paper for more information about the 3:07 model is in the description i hope you 3:09 enjoyed this short video i just saw this 3:12 news when i woke up and i couldn't do 3:13 anything else than make this video 3:15 before starting my day it's just too 3:17 exciting i will see you next week with another amazing paper

Deepmind May Have Just Created the World's First General AI

Deepmind puede haber creado la primera IA general del mundo

Deepmind vient peut-être de créer la première IA générale au monde

दीपमाइंड मे ने हाल ही में दुनिया का पहला जनरल एआई बनाया है

Deepmind は世界初の汎用 AI を作成した可能性があります

Deepmind pode ter acabado de criar a primeira IA geral do mundo

Deepmind có thể vừa tạo ra AI chung đầu tiên trên thế giới

Deepmind 可能刚刚创造了世界上第一个通用 AI

DeepMind just released a Generative model able to outperform widely-used nowcasting methods in 89% of situations for its accuracy and usefulness assessed by more than 50 expert meteorologists! Their model focuses on predicting precipitations in the next 2 hours and achieves that surprisingly well. It is a generative model, which means that it will generate the forecasts instead of simply predicting them. It basically takes radar data from the past to create future radar data. So using both time and spatial components from the past, they can generate what it will look like in the near future. You can see this as the same as Snapchat filters, taking your face and generating a new face with modifications on it. To train such a generative model, you need a bunch of data from both the human faces and the kind of face you want to generate. Then, using a very similar model trained for many hours, you will have a powerful generative model. This kind of model often uses GANs architectures for training purposes and then uses the generator model independently. If you are not familiar with generative models or GANs, I invite you to watch one of the many videos I made covering them, like . If this sounds either interesting or too good to be true, watch the video! this one about Toonify References ►Read the full article: ►Reference: Ravuri, S., Lenc, K., Willson, M., Kangin, D., Lam, R., Mirowski, P., Fitzsimons, M., Athanassiadou, M., Kashem, S., Madge, S. and Prudden, R., 2021. Skillful Precipitation Nowcasting using Deep Generative Models of Radar, ►Colab Notebook: ►My Newsletter (A new AI application explained weekly to your emails!): https://www.louisbouchard.ai/deepmind-rain-nowcasting/ https://www.nature.com/articles/s41586-021-03854-z https://github.com/deepmind/deepmind-research/tree/master/nowcasting https://www.louisbouchard.ai/newsletter/ Video Transcript 00:00 You've most certainly planned a trip to the beach for the day, checked the weather before 00:04 going, which said it was sunny, and just when you arrived, it started raining. 00:08 This or a similar situation happened to all of us. 00:11 We always talk about the weather. 00:13 There are two big reasons for that: it has a big impact on our lives and activities, 00:18 and we sometimes have nothing better to talk about. 00:20 A common agreement is that it seems like weather forecasts in the next few hours are completely 00:26 random. 00:27 Especially when it comes to rain prediction. 00:29 Well, there's a reason for that. 00:31 It's actually highly complex. 00:33 These short-term weather predictions are called Precipitation Nowcasting and are made using 00:38 various methods to predict what will happen in the next two hours. 00:42 These methods are driven by powerful numerical weather prediction systems predicting the 00:46 weather by solving physical equations. 00:49 They are quite powerful for long-term predictions but struggle to find fine-grain forecasts 00:55 above your head at a specific time of the day. 00:57 It's just like statistics. 00:58 It is easy to predict what an average human will do in a situation but quite impossible 01:03 to predict what a particular individual will do. 01:06 If you'd like to get deeper into these mathematical models, I already explained how they work 01:10 in more detail in my other video about global weather prediction using deep learning. 01:14 Even though we have a lot of radar data to predict what will happen, the mathematical 01:19 and probabilistic-based methods fail to be precise. 01:22 You can see where this is going. 01:24 When there's data, there's AI. 01:25 Indeed, this lack of precision may change in the future, and in part because of DeepMind. 01:29 DeepMind just released a Generative model able to outperform widely-used nowcasting 01:35 methods in 89% of situations for its accuracy and usefulness assessed by more than 50 expert 01:42 meteorologists! 01:43 Their model focuses on predicting precipitations in the next 2 hours and achieves that surprisingly 01:49 well. 01:50 As I just said, it is a generative model, which means that it will generate the forecasts 01:54 instead of simply predicting them. 01:56 It basically takes radar data from the past to create future radar data. 02:00 So using both time and spatial components from the past, they can generate what it will 02:05 look like in the near future. 02:07 You can see this as the same as Snapchat filters, taking your face and generating a new face 02:12 with modifications on it. 02:14 To train such a generative model, you need a bunch of data from both the human faces 02:19 and the kind of face you want to generate. 02:21 Then, using a very similar model trained for many hours, you will have a powerful generative 02:27 model. 02:28 This kind of model often uses GANs architectures for training purposes and then uses the generator 02:34 model independently. 02:35 If you are not familiar with generative models or GANs, I invite you to watch one of the 02:39 many videos I made covering them, like the one appearing on the top right corner right 02:44 now. 02:45 One of the most basic architectures to achieve image generation is called a UNet. 02:49 It basically takes an image, or past radar data, in this case, encodes it using trained 02:54 parameters, and takes this encoded information to generate a new version of the same image, 03:00 which in our case would be the same radar data of the next few minutes. 03:05 Here's what it looks like when feeding a typical UNet with forecast data compared to what it 03:09 should look like, the target. 03:12 You can see that it is relatively good, but not really precise and surely not enough to 03:16 be used in our daily lives. 03:18 Here's a comparison with the currently used numerical weather prediction approach like 03:22 PySTEPS. 03:23 It's a bit better, but you can see that it is not perfect either. 03:27 We cannot really further improve the probabilistic methods using math equations, so trying different 03:32 approaches becomes interesting. 03:34 Also, the fact that we have a lot of radar data to train our models is quite encouraging 03:38 for the deep learning approaches. 03:40 This is why DeepMind successfully created a GAN-like architecture made explicitly for 03:45 this task. 03:46 And here are the results. 03:47 You can see how much closer it is to reality with more fine-grain details. 03:52 Really impressive! 03:53 They achieved that by using both time and spatial components from the past radar data 03:58 to generate what the radar data could look like in the near future. 04:02 By the way, if you find this interesting, I invite you to subscribe and like the video 04:06 and share the knowledge by sending this video to a friend. 04:09 I'm sure they will love it, and they will be grateful to learn something new because 04:13 of you! 04:14 And if you don't, no worries, thank you for watching! 04:16 So more precisely, the previous 20 minutes radar observations are sent to the model to 04:21 generate 90 minutes of possible future predictions. 04:25 It is trained just like any GAN architecture guiding the learning process by penalizing 04:29 the difference between the generated radar predictions and the real predictions, which 04:34 we have in our dataset for training. 04:37 As you can see here, there are two losses and a regularization term, which are the penalties 04:41 that will lead our model during training. 04:44 The first one is a temporal loss. 04:46 This temporal loss will force the model to be consistent in its generation over multiple 04:50 frames by comparing them with the real data over a specific amount of time (or frames) 04:56 to be temporarily realistic. 04:58 This will remove weird jumps or inconsistency over time that couldn't happen in the real 05:04 world. 05:05 The second loss is the same thing but for spatial purposes. 05:08 It ensures spatial consistency by comparing the actual radar data versus our generated 05:13 prediction at a specific frame. 05:15 In short, it will force the model to be "spatially intelligent" and produce confident predictions 05:20 instead of large blurry predictions like we saw for UNet. 05:24 The last term and not least is the regularization term. 05:27 It will penalize differences in the grid cell resolution between the real radar sequences 05:33 and our predictions using many examples at a time instead of comparing the predictions 05:38 one by one like the two losses. 05:40 This will improve performance and produce more location-accurate predictions. 05:45 So you will send some observations, get their predictions, compare them with the real radar 05:50 data you have with these three measurements we just covered, and update your model based 05:55 on the differences. 05:56 Then, you repeat this process numerous times with all your training data to end up with 06:01 a powerful model that learns how the weather is changing and can ideally accurately generalize 06:06 this behavior to most new data it will receive afterward. 06:10 You can see how accurate the results are where UNet is accurate but blurry, and the numerical 06:15 method overestimates the rainfall intensity over time. 06:19 Of course, as they say, no method is without limitations, and theirs struggle with long-term 06:24 predictions, and just like most deep learning applications, they also struggle with rare 06:29 events that do not frequently appear in the training datasets, which they will work on 06:33 improving. 06:34 Of course, this was just an overview of this new paper attacking the super interesting 06:38 and useful task of precipitation nowcasting. 06:42 I invite you to read their excellent paper for more technical detail about their implementation, 06:46 training data, evaluation metrics, and the expert meteorologist study. 06:50 They also made a Colab Notebook you can play with to generate predictions. 06:54 Both are linked in the description below. 06:57 Thank you very much for watching for those of you who are still here, and I will see 07:00 you next week!

AI Endorsed by Expert Meteorologists: DeepMind's Weather Forecast Model

DeepMind may allude to two things: the innovation behind Google’s man-made reasoning (AI) venture, and the organization that is liable for it. The organization DeepMind is an auxiliary of Alphabet, the parent organization of Google. DeepMind’s man-made reasoning innovation has discovered its way into an assortment of Google ventures and gadgets. In the event that you use Google Home or Google Assistant, at that point your life has just crossed with DeepMind here and there. Why Did Google Acquire DeepMind? DeepMind was established in 2011 with the objective of “understanding knowledge, and afterward utilizing that to comprehend everything else.” The originators handled the issue of AI utilizing bits of knowledge from the field of neuroscience. Their objective was to make incredible broadly useful calculations that would have the option to learn and reinvent themselves, instead of waiting be physically modified by people. A few players in the field of AI were dazzled by the ability of the DeepMind group. In 2012, Facebook made a play to gain the organization. That arrangement self-destructed, yet Google plunged in and obtained DeepMind in 2014 for about $500 million. DeepMind then turned into an auxiliary of Alphabet during the Google corporate rebuilding of 2015. Google’s fundamental explanation behind purchasing DeepMind was to kick off its man-made reasoning project. While DeepMind’s principle grounds stayed in London, England, an applied group was dispatched to Google’s base camp in Mountain View, California. That group was to take a shot at coordinating DeepMind AI with Google items. What Is Google Doing With DeepMind ? DeepMind’s objective of fathoming knowledge didn’t change when it gave the keys over to Google. Work proceeded on profound realizing, which is a kind of general AI program. This contrasts and prior AIs like the Deep Blue PC, which in 1996 broadly crushed chess Grandmaster Gary Kasparov. Such PCs exceeded expectations at area explicit errands however were insignificantly valuable outside of those areas. DeepMind, then again, was intended to gain as a matter of fact. DeepMind’s man-made reasoning has figured out how to play computer games like Breakout superior to the best human players. In 2016, a DeepMind-controlled program called AlphaGo crushed a title holder Go player—an achievement for the way that Go is significantly more confused than chess. Notwithstanding unadulterated exploration, Google has incorporated DeepMind AI into its leader search and cell phones, including Google Home and Android. How Google DeepMind Affect Your Life Daily ? DeepMind’s profound learning instruments have been executed over the whole range of Google items and administrations. In the event that you use Google, there’s a decent possibility you have cooperated with DeepMind here and there. Probably the most unmistakable uses for DeepMind AI incorporate discourse acknowledgment, picture acknowledgment, misrepresentation discovery, spam ID, penmanship acknowledgment, interpretation, Google Maps Street View, and Local Search. Google’s Super-Accurate Speech Recognition Discourse acknowledgment, or the capacity of a PC to decipher spoken orders, has been around for quite a while. Menial helpers like Siri, Cortana, Alexa, and Google Assistant have carried the usefulness closer to our every day lives. On account of Google’s voice acknowledgment innovation, profound learning has been conveyed to incredible impact. AI has permitted Google’s voice acknowledgment tech to accomplish a great degree of exactness for the English language, to where it is as precise as a human audience. In the event that you have any Google gadgets, similar to an Android Phone or Google Home, this directly affects your life. Each time you state, “OK, Google” trailed by an inquiry, DeepMind utilizes its muscles to help Google Assistant comprehend what you are stating. In contrast to Amazon’s Alexa, which utilizes eight receivers to comprehend voice orders, Google Home’s DeepMind-fueled voice acknowledgment framework just requires two. Google Home and Assistant Voice Generation Conventional discourse blend utilizes something many refer to as concatenative content to-discourse (TTS). At the point when you associate with a gadget that utilizes this strategy for discourse union, it counsels a database brimming with discourse parts and gathers them into words and sentences. This outcomes in an unusually curved discourse example, and it is normally evident that the speaker isn’t human. DeepMind handled voice age with a venture called WaveNet, which was intended to make misleadingly created voices sound increasingly common. WaveNet depends on tests of genuine human discourse however doesn’t utilize the examples to blend new voices. Rather, it investigates the examples of human discourse to figure out how the crude sound waveforms work. This permits the program to communicate in various dialects, use complements, or even be prepared to seem like a particular individual. In contrast to different TTS frameworks, WaveNet creates non-discourse seems as though breathing and lip-smacking to render a considerably progressively sensible vocal profile. In the event that you need to hear the distinction between a voice created through concatenative content to-discourse, and one produced by WaveNet, DeepMind makes them intrigue voice tests that you can tune in to. Profound Learning/Deep learning and Google Image Search Without man-made reasoning, looking for pictures depends on setting hints like labels, close by text, and record names. With DeepMind’s profound learning instruments, Google Image Search had the option to realize what different individuals and items resemble, permitting you to look through your own pictures and get significant outcomes without expecting to label anything. For instance, on the off chance that you scan for “hound,” Google will pull up pictures of your canine that you took, regardless of whether you never named them. This is on the grounds that it had the option to realize what pooches resemble, similarly that people realize what things resemble. Also, not at all like Google’s pooch fixated Deep Dream, it’s in excess of 90 percent exact at distinguishing a wide range of various pictures. DeepMind in Google Lens and Visual Search One of the most shocking improvements of DeepMind is Google Lens, a visual web crawler that permits you to snap a photo of an article in reality and right away draw up data about it. While the execution is unique, this is like how profound learning is utilized in Google Image search. At the point when you snap a photo, Google Lens can take a gander at it and make sense of what it is. In light of that information, it would then be able to play out an assortment of further developed activities. For instance, in the event that you snap a photo of a renowned milestone, it will gracefully you with data about the milestone. In the event that you snap a photo of a nearby store, it can pull up data about that store. In the event that the image incorporates a telephone number, Google Lens can perceive the data and give you the choice of calling the number.

Learn About Google DeepMind –

Will DeepMind use GANs to write the next Harry Potter?

DeepMind's AlphaZero Engineering Use Cases

This week I had the honour to talk with Petar Veličković, a research scientist at and affiliate lecturer at the University of Cambridge. In this captivating discussion, Petar shares insights into his academic background, his transition into machine learning, the value of pursuing a PhD in the field, and much more, including building Map’s travel time prediction algorithm! DeepMind Google Petar's journey began in Serbia, where he attended a mathematics-oriented high school. With a curriculum focused on mathematics, computer science, and physics, he developed a strong foundation in these subjects. Later, he pursued his undergraduate degree in computer science and his PhD in machine learning at the University of Cambridge. Petar's initial interest was in programming, particularly classical algorithms and competitive programming. However, through his academic journey, he discovered his passion for research. A software engineering internship helped him realize that research allowed him to tackle unsolved problems and make innovative contributions. This realization led him to explore research-oriented internships and ultimately pursue a PhD in machine learning. During his PhD, Petar embraced the opportunity to delve deep into research and push the boundaries of his field. He emphasizes that a successful PhD goes beyond becoming an expert in a specific area—it teaches adaptability and the ability to explore new directions. He also highlights the importance of networking, collaboration, and the lasting connections he made during his doctoral studies. Addressing the concern of specialization, Petar emphasizes that a PhD does not limit one's future prospects. It serves as an entry ticket, demonstrating the ability to persist and create work towards a research objective. He emphasizes that research trends change rapidly, and a PhD equips individuals with the skills to adapt and explore various avenues. Moreover, Petar debunks the misconception that a PhD is a requirement to excel in the industry. While a PhD offers valuable experience for research scientists, engineers, and other similar roles, it is not necessary for everyone. The field of AI welcomes individuals from diverse backgrounds and experiences. With the low barrier to entry and the availability of specialized knowledge, one can engage in cutting-edge research without formal machine learning or computer science education. machine learning Petar Veličković provides valuable insights into his academic journey, the transformative nature of a PhD, and the evolving landscape of AI research. His experiences serve as an inspiration for aspiring researchers, showcasing the importance of curiosity, adaptability, and the pursuit of innovative solutions. To gain further knowledge and insight, be sure to listen to the complete interview with Petar Veličković on , , or . Spotify Apple Podcasts YouTube Explore Petar's fascinating path and gain a deeper understanding of the world of AI research at such a great company like Deepmind. Discover the challenges he faced, the discoveries he made, and the impact he envisions for the future (e.g. AGI!). Watch the Video https://youtu.be/l2cj0-HTCC4?embedable=true&transcript=true

An Interview with Google Map's Time Prediction Algorithm Creator

Una entrevista con el creador del algoritmo de predicción de tiempo de Google Map

Une entrevue avec le créateur d'algorithmes de prédiction de temps de Google Map

Google मानचित्र के समय पूर्वानुमान एल्गोरिथम निर्माता के साथ एक साक्षात्कार

Google マップの時刻予測アルゴリズム作成者へのインタビュー

Uma entrevista com o criador do algoritmo de previsão de tempo do Google Maps

Một cuộc phỏng vấn với người tạo ra thuật toán dự đoán thời gian của Google Map

谷歌地图时间预测算法创建者访谈

On December 27th, 2022—just a couple of days after Christmas and less than a month post-ChatGPT’s grand reveal to the world—a team of Google and DeepMind researchers published on a new large language model they named and its in a medical and clinical setting. their findings MedPaLM “encouraging results” And while the globe is still firmly grasped in a ChatGPT frenzy, its virtually anonymous “cousin” MedPaLM could potentially herald a not-too-distant revolution in how we administer care. So What Exactly Is MedPaLM? To understand what MedPaLM is, let’s first break down what each letter in its abbreviated name means (a bit of an acrostic poem, if you will): Med: Medical MedPaLM is based on Google’s 540 billion parameters strong and was specially adapted and optimized to be tested on , a newly formulated benchmark for testing large language models (LLMs) for medical and clinical applications combining six existing open question-answering datasets spanning professional medical exams, research, and consumer queries (NedQA, MedMCQA, PubMedQA, LiveQA, MedicationQA, and MMLU) as well as a new dataset named HealthSearchQA focused on general medical knowledge searched for by consumers. PaLM model MultiMedQA Pa: Pathways Pa stands for , a new AI architecture launched by in October 2021, designed to handle a multitude of tasks simultaneously and learn new tasks at hyper-speed. Breaking away from the conventional method of training machine learning models to perform a single task well, Pathways ushered in a new era of AI models capable of performing multiple tasks at once while also drawing upon and combining their existing skills to learn new tasks faster and more effectively. Pathways Google - Jeff Dean, Google Senior Fellow and SVP, Google Research. “We want [this] model to have different capabilities that can be called upon as needed and stitched together to perform new, more complex tasks – a bit closer to the way the mammalian brain generalizes across tasks.” ‍ Google's Pathways AI architecture ‍ LM: Language Model LM is short for (in this instance, a Large Language Model), a computational model designed to understand and generate human language. Or, in simpler terms, a statistical model that predicts the probability of the next word in a given sequence of words. Language Model Given a sentence such as "The cat sat on the ____," a language model can predict that the most likely word to follow is "mat," "rug," or "chair," based on the probabilities of these words appearing in the context of the sentence. Large language models can be trained on vast amounts of text corpora using machine learning algorithms to learn the patterns and structure of language. Once trained, these models can be utilized for a range of natural language processing tasks, including text generation, translation, sentiment analysis, and more. How Was MedPaLM Developed? MedPaLM is the product of an extensive iterative experiment conducted by a team of Google and researchers to assess the efficiency and accuracy of large language models for medical and clinical applications. DeepMind ‍ What does that mean? Imagine you’re trying to perfectly tune a violin. To hit the perfect note, you need to continuously increase and decrease string tension and turn the knobs ever so gently to reach the desired outcome. In many ways, this is precisely what MedPaLM’s creators did to achieve their objective. So rather than thinking about it as an entirely new language model, think about it as Google’s PaLM model deliberately and meticulously calibrated for maximum precision in a medical setting. Ok, so how was it “tuned”? To tune or, more precisely, train MedPaLM, its researchers used prompts. Prompts for large language models are pieces of text used to guide or direct the generation of a new text by the model. They are intended to help the model produce output that is relevant to a particular task or topic and can take various forms, such as fill-in-the-blank statements, questions, or context-setting phrases. So, for example: "Complete the following sentence: The capital of France is ____." "What is the best way to prepare a grilled cheese sandwich?" "Write a paragraph describing the causes of the American Civil War." "In this passage from a novel, a character is introduced: 'She was tall and thin, with piercing blue eyes and a serious expression on her face.' What is the character's name and what is she doing in this scene?" Why do I need a healthcare-purposed language model to tell me the best way to prepare a grilled cheese sandwich? You don’t. This is exactly why MedPaLM’s developers employed a technique called Instruction Prompt Tuning (IPT).  Instruction Prompt Tuning is a method of fine-tuning a large language model for a specific task by providing examples of desired input and output pairs in the form of prompts. So rather than construct an LLM with general answers to general topics, MedPaLM’s team used guidelines and exemplars from a panel of qualified clinicians for each consumer medical question-answering datasets to prompt tune Flan-PaLM (MedPaLM’s previous iteration) and form a hyper-healthcare-focused LLM: MedPaLM. ‍ ‍ So How’s MedPaLM Performing? Well, that really depends on what you want it to do. In terms of providing answers to consumer medical questions, 94.4% of MedPaLM’s answers were judged as directly addressing the user’s question intent—only 1.5% less than a human clinician. 80.3% of its answers were found to be helpful, 11% less than a human clinician but 30% better than its predecessor Flan-PaLM. Impressive? You bet. ‍ ‍ But in other fields tested on, MedPaLM leaves much to be desired. In 16.9% of its answers, MedPaLM retrieved incorrect information as opposed to 3.6% for human clinicians. Incorrect reasoning was seen in 10.1% of MedPaLM’s answers and in just 2.1% of clinician answers. Incorrect comprehension occurred in 18.7% of cases for MedPaLM and in 2.2% for clinicians. What Does This Mean for Healthcare? Trying to definitively predict how MedPaLM and other LLMs will be harnessed for clinical and medical purposes is a bit of a fool's game at this stage. As impressive as recent advancements in this field are, a single error could have disastrous results within the life-or-death stakes of medical care. One incorrect answer to a patient's query (i.e., what drug should be taken to address a specific condition) could lead to hospitalization and even death. This issue is further exacerbated by the fact that large language models are essentially enormous black boxes and lack the explainability or the ability to pinpoint exactly why they gave a certain input. Solving an error starts with knowing how to find the problem, and as spectacularly capable as MedPaLM may be, that's just one (out of many) things it can't do. So, What Can LLMs be Used for in Healthcare? Approximately 7,287 articles ( ) are published monthly in medical journals. To theoretically keep up with the literature, a physician would need to dedicate 627.5 hours per month, or about 29 hours per weekday, or 3.6 full-time equivalents of physician effort to the task. And though this isn't realistic in any scenario, LLMs are incredibly effective in summarizing, analyzing, and presenting key points from long-form text, drastically cutting down the time needed to stay up to date with the latest research. Summarizing and analyzing research papers: with an average of 5.16 minutes reading time Large language models can be used to improve the quality of clinical documentation by identifying missing or incomplete information in medical records. Improving clinical documentation: By generating appropriate codes for procedures and treatments, LLMs could help reduce errors and ensure clinicians are adequately compensated for their work. Coding and billing: Using LLMs, clinicians can identify trends and patterns in patient health and develop targeted interventions to improve population health outcomes. Population health management: By analyzing large amounts of data and identifying patterns and trends that may not be easily traceable with traditional statistical methods, LLMs can be employed to facilitate clinical research at increased velocity and scale. Clinical research: LLMs can be leveraged to sift through and analyze patient feedback and sentiment through thousands of surveys and social media interactions to identify critical areas for improvement in patient satisfaction, access, and experience. Patient satisfaction analysis: What About Appointment Scheduling? LLMs are only as good as the data they’ve been trained on. ChatGPT, for example—albeit its objectively impressive knowledge retrieval capabilities—was only trained on parameters dating to 2021. Any event that took place at any time past its cut-off date of September 2021 is entirely out of its scope. So even though Boris Johnson resigned as Prime Minister of the United Kingdom in September 2022 and has since been succeeded by two prime ministers (Liz Truss and currently Rishi Sunak), ChatGPT is (self-admittedly) completely unaware. ‍ ‍ But what if I train an LLM on a list of all of my health system’s physicians? Wouldn't that work? It would. But only until the moment even the most minute change takes place. So, for instance, if just one of your physicians moves to a different clinic, adds another specialty to their care offerings, updates their working hours, or even resigns, your appointment-scheduling LLM will provide your patients with wrong and misleading information. If even a single piece of data is no longer relevant, the entire model becomes corrupted and, therefore, obsolete. The same can be said for any use case requiring real-time data to perform accurately. So, for example, site search, prescription refills, and up-to-date FAQs. ‍ ‍ Oh, so LLMs can really only help me with administrative tasks? On their own, yes, which is nothing to sneeze at and will save healthcare practitioners innumerable hours otherwise spent tackling bureaucratic office work. But for any use cases that demand real-time, up-to-date data, LLMs can only serve as an ingredient of a much more adaptive and flexible conversational platform, not the entire dish. , Hyro’s CEO lays out the ingredients needed to mix LLMs with enterprise-level conversational AI platforms to create truly elastic patient interactions that are driven by continuously updated organizational knowledge (proprietary data) and dynamic business logic (ever-changing internal and external processes). In this in-depth analysis Israel Krush Big Hero 6 When asked in a recent what he thought of ChatGPT and other LLMs, Andy Chu, SVP of Product and Technology at Providence Digital Innovation Group, said the industry is still a a 2014 Disney film depicting a futuristic healthcare robot—But added he envisions this technology used to answer patient-related administrative, "decision-tree", or general health-education questions. Becker's Hospital Review interview "long way off from the days of Big Hero 6,” https://youtu.be/jbKSvOfqxF0?embedable=true Chu's thoughts echo a broader sentiment shared by most digital health leaders that LLMs such as MedPaLM should be greeted with equal parts of excitement and caution. While there's no question these extraordinary "machines" can automate and accelerate a vast spectrum of primarily administrative applications, there are still glaring red flags when it comes to medical and clinical use. Furthermore, LLMs' absence of data adaptability renders them ineffective for any goal which calls for current and evolving internal organizational knowledge. Chu put it best by saying: "It's important to remember that healthcare is very personal, and generative AI technologies are as good as the data accessed. Data sources within healthcare are rightfully secure and protected, so we must be careful not to overhype the promise — especially when it comes to areas such as symptom detections, triage, and clinical treatment suggestions." Also published here.

An Intro to MedPaLM: ChatGPT's Healthcare-Focused "Cousin" 

Artificial General Intelligence (AGI) is one of the most important research topics humans can work on. It is one of the most hyped, speculated, feared, discussed topics in Artificial Intelligence (AI) research. AI is a very broad topic that in our opinion has a lot of boring subfields, such as search, logic, probabilistic methods, classifiers and statistical methods. With all the advances of GPTs and GANs, AGI remains a hard problem to solve. At its core, general intelligence is hard to define and maybe impossible to achieve. While the brightest minds at Google’s Deepmind and Open AI are working hard exploring ways to solve AGI. It seems that many researchers miss important concepts that I believe are essential to solving general intelligence. Numenta’s researchers recently outlined four basic concepts that I agree are vital for achieving AGI; these ideas are worth considering in broad AI research. 1. Continuous Learning When ML algorithm training happens, it is done on a static dataset, and when the training process is concluded, the algorithm can then be used. For example, when Open AI GPT3 is trained in 2021 it may be aware of COVID19, however, it will not be aware of the ongoing tensions in Ukraine, it will only know about Ukraine once it is trained on 2022 internet text data. You, on the other hand, know about tensions in Ukraine. Why is that? The answer is Continuous Learning. Another example, when Tesla's Autopilot got trained for rolling stops, it kept performing rolling stops until it was updated to stop this feature, not because it learned that this is wrong. The point is: the human brain keeps learning and updating its model of the world continuously, current AI algorithms don’t. Continuously learning AI algorithms that work will be a breakthrough towards AGI. Tesla kept performing rolling stops until it was updated to stop this feature, not because it learned that this is wrong, image: Forbes 2. Physical World Learning and Exploration General intelligence, at the first glance, seems to have little relevance to physics or the physical world. However, for an AI algorithm to make generally intelligent decisions or behave in an intelligent way in the human physical world. It has to have the ability to experience and experiment with real-world physics. Otherwise complex reasoning, decisions, or world related action performed by the AI may result in sporadic behavior. To add more to this, I believe that the AI algorithm should not be trained in a simulated virtual world since humans don’t fully understand the physical world. In a nutshell, the human brain is believed to learn about the physical world via movement. Then a breakthrough AI algorithm will learn the laws of physics by moving in the physical world. Boston Dynamics Atlas carrying a box, Image: Gizmodo 3. Generalization Zero Shot Learning (ZSL) happens when a child sees a new car that she/he had never seen before they know that this is a car and react accordingly. This generalization could be based on previous learning or intuition I believe has to do with the model-free methods the brain may be using. A breakthrough AI will have a certain structure that allows a degree of generalization or extrapolation without catastrophic results. When a child sees a new car that she/he had never seen before they know that this is a car and can react accordingly, Image: Pexels 4. Reference Frame Learning Jeff Hawkins’s 1000 brain theory of intelligence is based on the concept of reference frames. As I understand it, reference frames are abstract concepts (connections between neurons) the brain creates and stores (maintains). According to the theory, the brain uses these reference frames to think, plan, and predict. A breakthrough AGI algorithm may have a similar structure to reference frames to map and store complex concepts. Current designs of ANN may allow the existence of such connections, however, a breakthrough would be the automation of forming, modification, and distraction of such connections based on continuous learning, physical world exploration and generalization. The brain uses these reference frames to think, plan, and predict, Image: Pixabay If AI researchers and technologists are successful in embracing these concepts, I expect that AI with human intelligent features may emerge. Let’s hope they are reading this. Thanks to Christy Maver, Donna Dubinsky and Subutai Ahmad at Numenta for sharing their valuable ideas about the Biological Approach to Machine Intelligence If you learned something new or liked this content follow me For questions, showing support, and feedback contact me here here

Stories, News & Mentions about DEEPMIND

Categories

Trending Topics

Stories, News & Mentions about DEEPMIND

Deepmind’s Stories on HackerNoon

Deepmind's News & Mentions

Categories

Trending Topics